Systematic removal of Z-ring modulating proteins

To define the minimal set of proteins required for cell division in B. subtilis, we set out to remove as many FtsZ-interacting proteins as possible. The natural genetic transformation ability of B. subtilis facilitates the combination of multiple genomic mutations, however the number of antibiotic resistance markers necessary for selection are limited and therefore we made use of an established marker-less gene deletion method [43]. In short, an IPTG-inducible mazF toxin cassette was integrated next to a cell division gene, and subsequent induction of the MazF toxin resulted in the excision of the toxin cassette together with the cell division gene due to intra-molecular homologous recombination. The mazF toxin cassettes were separately introduced into the 168 wild type strain, and chromosomal DNA from these single insertion strains was used to construct strains with multiple deletions using natural genetic transformation. This procedure was chosen since chromosomal DNA integrates with much higher efficiencies compared to plasmid DNA. This strategy can be important as cell division defects can affect natural transformation. Deletion strains were labelled BMD (Bacillus Minimal Divisome) and they were grown in LB medium enriched with 1% glucose and 10 mM Mg²⁺. The reason for these additions was that some deletion strains grew slower in LB medium, which could be mitigated by adding glucose. Extra Mg²⁺ (10 mM) was added because of the ugtP deletion that we wanted to include (Fig 1) [44]. UgtP is required for synthesis of lipoteichoic acids, and the absence of this cell wall component results in bulging, which can be prevented by the addition of extra Mg²⁺ in the medium [44].

We have tried different combinations of successive gene deletions, and some were more successful than others (Fig A in S1 Appendix). Eventually, we succeeded in deleting 8 cell division genes, as shown in Fig 2A. We started out with a ∆zapA deletion mutant (BMD1), since this mutant by itself has no clear phenotype [22]. Next, we removed the genes coding for the negative regulators MinC and UgtP (BMD2 and BMD3, respectively). This enabled us to remove the genes minJ and ezrA (BMD5 and BMD6, respectively). These multiple deletions had only a mild effect on either growth or cell length (Fig 2B and 2C). This in itself was surprising since it has been reported that a ∆zapA ∆ezrA double deletion grows very slowly and is filamentous [22]. Inactivation of the negative regulator ClpX was more complicated since this protein is also required for the development of genetic competence in B. subtilis. However, this requirement can be bypassed by removing the pleiotropic transcriptional regulator Spx [45]. Therefore, spx was deleted prior to clpX (BMD7 and BMD9, respectively). When we subsequently tried to delete noc (BMD12), only a few transformants were obtained, and a PCR check revealed that, although noc was absent in these transformants, the ezrA gene was restored due to a second double crossover event, which was possible since we used chromosomal DNA from a strain that contains only a noc deletion. The repair of ezrA was in a way not surprising since it has been shown that a ∆noc ∆ezrA double mutant is very sick [42]. Further attempts to delete one of the remaining cell division genes (ezrA, sepF, ftsA) using the marker-less method were unsuccessful. To be sure that it was indeed not possible to delete these genes, we tried to remove them by using chromosomal DNA from strains in which these genes were replaced by an antibiotic resistance marker. Interestingly, using this approach, we were able to remove subsequently ezrA and ftsA, resulting in BMD14 and BMD27, respectively. Deletion of these genes slightly reduces the growth rate and resulted in very long cells, comparable to a single ftsA mutant (Fig 2B and 2C). BMD27 lacked MinJ and all the main FtsZ-interacting proteins shown in Fig 1, except for SepF. Despite several attempts, we were unable to delete sepF in either BMD12 or BMD14.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Stepwise deletion of 8 cell division genes.

(A) Schematic overview of the sequential deletion of 8 conserved cell division proteins in B. subtilis. Of note, spx is not related to cell division, however a spx deletion restores genetic competence in a ΔclpX mutant. The resulting mutant strains were labelled BMD. See main text for more details. (B) Growth curves of BMD strains grown in microtiter plates at 37⁰C in LB medium supplemented with 1% glucose and 10 mM MgSO₄. Insert highlights the growth curves of wild type, BMD27, and the single ftsA::erm deletion strains. Graphs are representatives of at least 3 biological experiments, the 2 others are shown in Fig B in S1 Appendix. (C) Cell length measurement of the BMD strains. 3 biological replicates, for each more than 400 cells measured. Mini cells were not included. Mean and standard deviations are indicated.

https://doi.org/10.1371/journal.pgen.1011567.g002

Phenotype of BMD27

To investigate whether BMD27 makes Z-rings, an ectopic xylose-inducible ftsZ-gfp reporter fusion was introduced (strain TNVS385). Fig 3A shows microscopic images of a representative cell expressing the FtsZ reporter fusion. The absence of 8 cell division proteins resulted in long filamentous cells with less Z-rings and polar deformation that seem to be caused by aberrant polar divisions (Fig 3B). It should be mentioned that in the stationary phase these cells became much shorter as they were apparently able to complete delayed division (Fig E in S1 Appendix). Except for the multiple polar minicells, these long cells strongly resemble an ftsA single deletion mutant (Fig 3C, strain TNVS553). In 39% of the TNVS385 cells FtsZ formed aberrant helical structures and long arcs (Fig 3A). In the ftsA::erm single deletion mutant this was seen for approximately 21% of the Z-structures. The Z-ring in Fig 3A seems to bisect a nucleoid. This occurred regularly, with a frequency twice that of the ftsA::erm single deletion mutant (~43% and ~22%, respectively). Nevertheless, more than half of the Z-rings assembled in between nucleoids, indicating that some form of nucleoid occlusion is still functional in the multiple deletions strain TNVS385.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Phenotype of BMD27.

(A) Fluorescence microscopy image of BMD27 cells expressing an ectopic FtsZ-GFP reporter fusion (green) (strains TNVS385). Membrane (red) and nucleoids (blue) were stained with FM5-95 and DAPI, respectively. FtsZ-GFP forms normal Z-rings (asterisk) and aberrant helical structures at cell poles. The Z-ring in this example seem to bisect a nucleoid. For a large field image of several cells see Fig C in S1 Appendix. (B) Examples of SIM images of BMD27 cell poles containing several minicells and aberrant Z-structures. (C) Fluorescence microscopy image of an ftsA::erm mutant expressing an ectopic FtsZ-GFP reporter fusion (strain TNVS553). The star marked Z-ring forms over a nucleoid. For a large field image of several cells see Fig D in S1 Appendix. Cells were grown to OD₆₀₀ ~ 0.5 at 30 °C in LB supplemented with 10 mM MgSO₄, 1% fructose, and 1% xylose to induce FtsZ-GFP. Fructose was used instead of glucose because of glucose-dependent catabolite repression of the xylose promoter [105]. Scale bars, 2 µm.

https://doi.org/10.1371/journal.pgen.1011567.g003

Previous studies have reported that the Min and nucleoid occlusion systems are not required for the positioning of Z-rings in the middle of cells in between the newly formed daughter chromosomes in B. subtilis and E. coli [46,47]. As shown in Fig 4, we found the same, and even in the multiple deletion strain BMD14 with an average cell length of more than 10 μm, there was still some preference for division at midcell (Fig 4). However, removing FtsA seems to abolish midcell selection, which was also the case for a ftsA single deletion mutant (Fig 4). Since these mutants forms cells with a comparable length as the BMD14 mutant containing FtsA (Fig 2C), it seems like FtsA has some regulatory role in FtsZ ring positioning.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Position of nascent cell division sites in deletion mutants.

The relative position of division sites was determined by staining cells with the fluorescent membrane dye FM5-95. The distance to one cell pole was measured and divided by the total cell length. To better indicate midcell preference, division positions were randomized relative to the left and right cell poles. Mean cell lengths are indicated between brackets. Data based on 3 biological replicates, minimally 75 cells were measured for the wild type and BMD1-12 strains, and minimally 50 cells for BMD14 and the ftsA::erm strain. Cells were grown in LB supplemented with 10 mM MgSO₄ and 1% glucose at 37 °C to exponential phase.

https://doi.org/10.1371/journal.pgen.1011567.g004

Whole genome sequencing

Since we were able to delete gene combinations that normally would have resulted in lethal phenotypes, it might be that suppressor mutations were collected during the process. To examine this, whole genome sequencing of all 10 deletion strains was performed. In total 3 silent mutations, 9 substitutions, 2 nucleotide insertions, and 8 deletions were found (Table 1), and their positions on the genome are schematically presented in Fig 5.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. List of spontaneous mutations in the genomes of the different BMD mutants. Multi-nucleotide and gene deletions are indicated by ∆, and a stop codon mutation by *.

https://doi.org/10.1371/journal.pgen.1011567.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Location of spontaneous mutations in the BMD genomes.

Schematic depiction of spontaneous mutations identified by whole genome sequencing on the 4.25 Mbp B. subtilis genome (X-axis). Deleted cell division genes are indicated as orange squares, and spontaneous mutations are indicated as green triangles. The rsgI-ogt deletion covers 10 genes. See main text for details.

https://doi.org/10.1371/journal.pgen.1011567.g005

The first mutations occurred when minC was deleted (BMD2), yielding a point mutation (M66R) in the proton/glutamate symporter GltP [48]. The subsequent deletion of ugtP (BMD3) did not result in any mutation, however the removal of minJ (BMD5) caused the deletion of the complete 22 kb long Integrative and Conjugative Element ICEBs1. It has been shown that the inactivation of ugtP increases ICEBs1 conjugation efficiency [49], which may have facilitated the excision of ICEBs1. Subsequent deletion of ezrA (BMD6) resulted in a 6-basepair deletion in the N-terminal domain of the membrane protease protein DdcP [50], which has recently been shown to degrade YneA, the inhibitor of cell division during the SOS response [51]. How this would suppress any cell division phenotype is unclear. Deletion of spx did not cause any spontaneous mutations, however, the subsequent deletion of clpX (BMD9) yielded a 1.6 kb long deletion containing three genes, mgsR, yqgY and yqgX, the latter two coding for unknown proteins. MgsR moderates the expression of a subset of SigB regulated stress genes and is controlled by proteolytic degradation by the ClpXP protease complex [52,53]. The increased stability of MgsR after removal of ClpX might have provided selection pressure favoring this 1.6 kb genomic deletion. In addition, BMD9 contains point mutations in uxuA and yoaN, coding for a mannonate dehydrogenase and an inner spore coat protein, respectively [54,55].

Subsequent deletion of noc (BMD12) restored the ezrA gene, due to an additional double crossover event, and in addition to this, the strain acquired several spontaneous mutations among which in ponA (G145R) and spoVG (E90*). Interestingly, the glycosyltransferase-transpeptidase PBP1, encoded by ponA and involved in lateral and septal peptidoglycan synthesis, is normally recruited to the cell division site by EzrA [20]. The conserved RNA binding protein SpoVG is a negative regulator of asymmetric cell division during sporulation [56]. In addition to these 2 mutations, BMD12 also obtained a triple amino acid deletion in the proline permease PutP (aa 356–358), and amino acid substitutions in the hypothetical proteins YuiF (L64F) and YwoF (G333V), the membrane protease SppA (A96V) [57], and a stop codon in the hypothetical protein YpmT (D31*).

When ezrA was again deleted in strain BMD14, mutations in braB (S343P), rapA (F186L) and sftA (P375S) were found. BraB is a branched-chain amino acid transporter [58], and RapA is a phosphatase involved in the quorum sensing regulation of sporulation [59]. The mutation in sftA is an interesting one since SftA is a conserved membrane-attached DNA translocase, with some homology to FtsK in E. coli. The protein binds to the Z-ring supported by both SepF and FtsA [60]. It has been speculated that the protein, like FtsK, prevents trapping of chromosomal DNA during septum synthesis by pumping DNA away from a closing septum. However, it is more likely that this function is executed by the FtsK homologue SpoIIIE in B. subtilis [60–63]. In E. coli, FtsK is an essential component of the divisome, but in B. subtilis, SftA is not required for cell division, although inactivation of the protein leads to slightly elongated cells [61,62,64]. It was shown that overexpression of SftA blocks cell division in an ezrA mutant, possibly because of interference with the assembly of SepF [60]. This activity may account for the spontaneous mutation in SftA. The disappearance of the gltP, ezrA and sppA mutations in BMD3, BMD12 and BMD14, respectively, can be explained by an extra double crossover event, as explained earlier. However, we have no explanation why the 2 bp deletion (GCCCA to GCC) in gltP in BMD 2 reappeared in BMD6 and again in BMD14.

Ultimately, removing ftsA (BMD27) yielded a large 6-kb deletion, covering 10 genes: rsgI, sspD, ykrK, htpX, ktrD, ykzP, ykzE, ykrP, kinE and ogt. RsgI is an anti-sigma factor controlling the activity of sigma I, which is required for the activation of general stress proteins, but also stimulates expression of the actin-homologue and cell shape-determining protein MreBH, and the undecaprenyl pyrophosphate phosphatase BcrC, involved in synthesis of the carrier lipid for cell wall synthesis [65]. Inactivation of RsgI has been shown to suppress certain cell shape defects in B. subtilis [66]. SspD is a protein found in spores [67], and KinE is a two-component sensor kinase involved in sporulation [68]. HtpX is a stress-responsive membrane protease whose expression is controlled by YkrK [69], and ktrD codes for a potassium transporter [70]. Lastly, the genes ykzP, ykzE and ykrP code for proteins with unknown functions, and the ogt gene codes for O6-methylguanine DNA alkyltransferase required for DNA-alkylation repair [71].

Suppressor mutations

To test whether the suppressor mutations might play a role in cell division, we constructed deletion strains of all mutated genes in the wild type background, except for those that were located in the ICEBs conjugation element, and measured the effect on cell length (Fig 6A). The most apparent results were the longer cells of the ponA and sftA mutants, which deletions are known to increase cell length [61,62,72], and the shorter cell length of the braB mutant.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Cell division characteristics of potential suppressor mutants.

(A) Cell length measurements of single deletion mutants. Colors indicate the different gene deletion step in which the related mutations appeared (deleted gene indicated above). Based on 3 biological replicates, at least 500 cells were measured for every mutant. Mean and standard deviations are indicated. (B) Spot-dilution assay to test the sensitivity of single deletion mutants for increased concentrations of the FtsZ inhibitor 3-MBA. (C) Spot-dilution assay to test the sensitivity of single deletion mutants for decreasing FtsZ concentrations. The mutants contained an IPTG-inducible ftsZ allele, and were spotted on plates with decreasing concentrations of IPTG. ∆zapA was included as positive control and wild type and a ∆kinE mutant as negative controls.

https://doi.org/10.1371/journal.pgen.1011567.g006

Mutations that affect cell division are more likely to show a clear phenotype when cell division is disturbed. This can, e.g., be achieved by treating cells with the FtsZ inhibitor 3-methoxybenzamide (3-MBA) [73]. To test whether the different deletion mutants showed any resistance to 3-MBA, the mutants were spotted onto agar plates containing increasing concentrations of 3-MBA. As shown in Fig 6B, the ponA and braB deletion mutants were clearly more resistant to 3-MBA, indicating that these are likely functional suppressor mutations in the BMD strains.

The efficacy of 3-MBA might be affected by changes in the cell envelope. Therefore, we assessed the results by testing the sensitivity of the most relevant deletion mutants, including ∆braB, ∆ddcP(ylbL), ∆ponA, ∆spoVG and ∆sftA, for reduced FtsZ levels, by introducing an IPTG-inducible ftsZ allele (Pspac-ftsZ). As a positive control we included a ∆zapA strain, which has been shown to be very sensitive for low cellular FtsZ levels [22], and ∆kinE as a negative control. The spot-dilution assay was performed on agar plates containing different IPTG concentrations. As expected, the zapA deletion was most sensitive (Fig 6C), whereas the braB mutant showed the strongest resistance to low IPTG concentrations (Fig 6C). It is not immediately clear why ∆ponA showed sensitivity towards lower FtsZ concentrations, whereas it resisted high 3-MBA concentrations.

Phenotype of a ∆braB mutant

Deletion of braB resulted in a slightly lower growth rate compared to the wild type strain (Fig F in S1 Appendix), but there is no clear reason why this would cause an increased resistance to 3-MBA and to lower FtsZ levels. However, braB is located adjacent to ezrA on the genome and reads against it (Fig 7A), so the effect of a braB deletion on cell division might be indirect and caused by a polar effect on the activity of ezrA. To exclude this possibility, we inserted a xylose-inducible braB-gfp fusion into the ectopic amyE locus. The fluorescent GFP reporter was added to follow the localization of BraB in the cell. As shown in Fig 7B, expression of the fusion protein restored the sensitivity for 3-MBA in a ∆braB background (strain TNVS308), indicating that the BraB-GFP fusion is active, and that the cell division effect of a braB mutation is unrelated to ezrA. The BraB-GFP signal covers the cell membrane and there was no strong accumulation at division sites (Fig 7C). The slight increase in fluorescence at cell division sites can be explained by the presence of a double cell membrane at these sites, and the position of these membranes parallel to the line of sight. The fluorescence pattern suggests that BraB does not regulate cell division directly.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. BraB-GFP localization.

(A) Schematic presentation of the braB genome location. (B) 3-MBA spot-dilution assay, showing that the sensitivity of a ∆braB strain (strain TNVS292) for the FtsZ inhibitor 3-MBA is restored by ectopic expression of a BraB-GFP fusion (strain TNVS308). (C) Fluorescence light microscopy showing membrane localization of BraB-GFP in ∆braB (strain TNVS308). All cells were grown in LB supplemented with 0.1% xylose at 37 °C. Samples were taken at exponential phase for microscopic analyses and the spot-dilution assay. Scale bar, 2 µm.

https://doi.org/10.1371/journal.pgen.1011567.g007

Effect of ∆braB on the cell membrane

It is unclear how BraB, a branched-chain amino acid (leucine, isoleucine, valine) transporter, can influence cell division. B. subtilis contains two other branched-chain amino acid permeases, BcaP and BrnQ, and it has been shown that the absence of braB does not affect the uptake of branched-chain amino acids [58]. We deleted both bcaP and brnQ, but neither mutations changed the sensitivity towards 3-MBA (Fig G in S1 Appendix). While measuring the cell length of ∆braB cells, we did notice that the fluorescent membrane dye FM5-95 gave an irregular fluorescent membrane stain (Fig 8A). Such irregular fluorescent membrane staining has been documented before and is presumed to be caused by local enrichment of more fluid lipids, i.e., lipids containing short, branched and/or unsaturated fatty acids. These domains can increase the concentration and/or fluorescence intensity of membrane probes and are also referred to as RIFs (regions of increased fluidity) [74]. Many bacteria contain two types of branched-chain fatty acids, iso- and anteiso-fatty acids, which differ in the position of the terminal methyl side chain. The iso form, with a methyl side chain at the 2^nd terminal C atom, requires final attachment of either valine or leucine to the carbon chain, whereas the anteiso form, with a methyl side chain at the 3^rd terminal C atom, requires isoleucine as final attachment to the carbon chain. Thus, inactivation of BraB might influence the concentration of these fatty acids. However, a fatty acid analysis of wild type and ∆braB cells did not show lower levels of branched-chain fatty acids in the mutant, although there was a clear shift from anteiso (fluid) to iso (less fluid) fatty acids (Fig 8B). On the other hand, the level of unsaturated and short-chain fatty acids, both more fluid, increased as well (Fig 8B). To determine whether these alterations affected the overall fluidity of the cell membrane, we performed a Laurdan fluorescence general polarization measurement. The fluorescence emission spectrum of Laurdan depends on lipid head group spreading and fatty acyl chain flexibility, making it a good indicator of membrane fluidity [75]. As shown in Fig 8C, the absence of BraB increased the overall membrane fluidity considerably.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. braB deletion affects membrane composition.

(A) Aberrant fluorescent membrane pattern in exponentially growing ∆braB cells compared to wild type cells. Cells were stained with the membrane dye FM5-95. Wide field images with more cells are shown in Fig F in S1 Appendix. Strains used: wt and TNVS292. Scale bars, 2 µm. (B) Ratio of the fatty acid composition of the membrane are presented as mean and standard deviation of duplicate measurements. Anteiso-, iso-, saturated and unsaturated fatty acids are indicated and color-coded white to dark grey for clarity, respectively. Long fatty acids contain 16 or more carbons and short fatty acids have less. Strains used: wt and TNVS292. (C) Ratios of straight (saturated and unsaturated) and branched (iso- and anteiso), saturated and unsaturated, iso and anteiso, and long and short chain fatty acids in wild type and ∆braB cells based on the data in table (B). Samples were taken at an OD₆₀₀ of 0.5. (D) Membrane fluidity of wild type and ∆braB cells measured with Laurdan fluorescence general polarization. Samples were taken at OD₆₀₀ of 0.2 and 0.8. As a control wild type cells were treated with the membrane fluidizer benzyl alcohol (BA). The laurdan analysis was performed with 3 biological replicates.

https://doi.org/10.1371/journal.pgen.1011567.g008

SepF and FtsA

Our data suggests that SepF is sufficient for the formation of an active Z-ring. SepF is conserved in Gram-positive bacteria and cyanobacteria and absent from other Gram-negative bacteria, where FtsA functions as the membrane anchor for FtsZ [12,13]. Many Gram-positives use both FtsA and SepF, like the Firmicutes, whereas, e.g., Corynebacteria only use SepF. Interestingly, archaea contain only SepF homologues and it was speculated that FtsZ and SepF may date back to the Last Universal Common Ancestor (LUCA) [15].

Cryo-EM of Caulobacter crescentus and E. coli has shown that FtsZ polymers are separated from the cell membrane by 15 to 16 nm [27,28]. This is a substantial distance that is caused by the presence of the 45 kDa membrane anchor FtsA binding to the flexible C-terminus of FtsZ [27]. It seems unlikely that late cell division proteins, which are all transmembrane proteins without large cytoplasmic domains, are able to bridge this distance and directly interact with FtsZ. In fact, for E. coli it has been shown that the recruitment of late cell division proteins relies on the interaction of FtsA with the cytoplasmic domain of the late cell division protein FtsN [77]. Gram-positive bacteria do not contain an FtsN homologue and it is unknown how the late cell division proteins are recruited to the Z-ring in these organisms. Our multiple deletion strain suggests that SepF, interfacing the Z-ring and cell membrane, is the link between early and late cell division proteins in Gram-positive bacteria. However, it is then likely that, at least in B. subtilis, this function is shared with FtsA, since a mutant lacking SepF can still divide [14].

SepF spontaneously forms strongly curved polymers with a diameter of approximately 43 nm, which is comparable to the thickness of the septal wall. In a previous study, we have shown that the specific curvature of SepF polymers determines the thickness of the division septum, likely by forming arc-like polymers that follow the leading edge of the nascent septal wall, creating precise septum wide scaffolds for FtsZ polymers and the associated proteins responsible for synthesis of the septal wall [78]. In the presence of ATP and a lipid bilayer E. coli FtsA can also form small ring-like polymers with a radius of about 20 nm [79]. However, there is no indication that this characteristic would somehow guide the width of the newly formed septal wall. This scaffold function of SepF might be the reason why we were unable to construct a minimal divisome mutant that uses only FtsA as membrane anchor for FtsZ.

Branched-chain amino acid transporter BraB

Whole genome sequencing revealed a number of spontaneous suppressor mutations among which one maps to braB. The S343P mutation in BraB is located in a putative transmembrane domain, likely disabling the protein. A braB deletion does not affect the uptake of branched-chain amino acids [58], and inactivation of the other branched chain amino acid encoding transporters, bcaP and brnQ, did not convert resistance to 3-MBA. It is unlikely that the protein interacts directly with one of the structural cell division proteins since a BraB-GFP fusion showed no preference for cell division sites. In many bacteria branched chain amino acids are incorporated at the end of fatty acids and are way to regulated membrane fluidity [80].

Role of lipids in cell division

It has long been speculated that membrane lipids play a role in bacterial cell division. More than 50 years ago Francois Jacob and co-workers showed that membrane synthesis occurs primarily in the middle of B. subtilis cells, and postulated that there must be an intricate relationship between DNA replication, cell division and membrane synthesis in bacteria [81]. In fact, different studies have shown increased lipid synthesis during cell division in Bacillus [82,83]. However, so far there is no indication for a regulatory link between lipid synthesis and cell division. Several years ago, a microscopy study suggested that the acyltransferase PlsX, which catalyzes an essential step in lipid synthesis, binds to the Z-ring [84], but this turned out to be an artefact [85,86]. Nevertheless, certain lipid species can affect cell morphology in bacteria. For example, a reduction in anionic phospholipids in E. coli resulted in filamentous cells [87,88]. It was shown that the membrane association of E. coli MinD, which recruits MinC to the cell membrane, is stimulated by anionic phospholipids but also by an increase in unsaturated acyl chains [89]. The latter observation was attributed to facilitated insertion of the C-terminal amphipathic helix of MinD, which functions as a membrane anchor. Insertion of such a relatively bulky helix in between phospholipid molecules will cost less energy when lipids are less densely packed and the membrane is in a more fluid state [90]. Since the membrane fluidity in a ∆braB mutant is increased it might be that this stimulates the attachment of FtsA and SepF to the cell membrane, as both proteins use an amphipathic helix as membrane anchor [14,17]. We have examined whether changing the membrane fluidity of B. subtilis affects its sensitivity to 3-MBA and depletion of FtsZ. Changing the membrane fluidity can be achieved by using a B. subtilis strain lacking the phospholipid desaturase gene des and the bkd operon, encoding the enzymes catalyzing the conversion of branched chain amino acids into intermediates for branched chain amino acid synthesis. This mutant can grow in minimal medium when supplied with either the iso- or anteiso-branched chain fatty acid precursors isobutyrate or 2-methyl butyrate, respectively. Feeding this mutant with the latter precursor will increase the membrane fluidity [91]. However, the presence of 2-methyl butyrate did not increase the resistance to either 3-MBA or depletion of FtsZ compared to ∆des ∆bkd mutant cells fed with isobutyrate (Fig H in S1 Appendix). Therefore, it seems unlikely that only a change a membrane fluidity accounts for the observed ∆braB phenotype.

Minimal divisome requirements

Our research was initiated by the question what minimal set of cell division proteins are needed for division of a bacterium that contains a cell wall. Clearly FtsZ, but this is not the case for cell wall-lacking bacteria. In their pioneering work, Hutchison and coworkers constructed a minimal cell by taking a relatively simple cell wall-lacking Mycoplasma mycoides cell and removing all non-essential genes [3]. Interestingly, this minimal cell did not require FtsZ, although its presence was shown to improve cell division and reduce cell size [92]. Some Mycoplasma species lack a ftsZ gene by nature, and it was already shown that the ftsZ gene of Mycoplasma genitalium can be inactivated without affecting viability [4]. Presumably, FtsZ is not vital for these bacteria because they lack a cell wall. In fact, B. subtilis and E. coli mutants that grow without a cell wall, so called L-forms, can propagate without FtsZ and produce daughter cells by spontaneous membrane blebbing [5,6]. There has been one example of a cell wall-forming bacterium that can propagate when its ftsZ gene is inactivated and that is the filamentous actinomycetes Streptomyces coelicolor. This bacterium grows by branching and primarily uses cell division for dividing hyphae into multiple sporulating cells. Although the latter was blocked in the ∆ftsZ mutant, cells were still able to form branching hyphae [93].

When designing a cell wall-containing synthetic bacterial cell, a mechanism must be in place to divide its cell wall, and our data suggests that FtsZ and SepF are sufficient for the positioning of proteins responsible for this, although we cannot exclude that some unknown protein might have become essential for cell division in the minimal divisome mutant. Interestingly, the minimal divisome mutant BMD27 resembles the situation found in most Actinobacteria (Streptomycetes, Mycobacteria, Corynebacteria etc), which all lack MinJ, ZapA, EzrA, FtsA and the Min and Noc systems, but which do contain SepF and FtsZ [94]. It is not yet clear how Actinobacteria control correct Z-ring placement at midcell [95], however even in BMD27 most divisions still occur in between nucleoids. Possibly the density of DNA polymers, maybe enriched close to the cell membrane by transertion (coupling of transcription, translation and secretion/membrane insertion), is sufficient to prevent efficient FtsZ polymerization over the nucleoids [47,96].

Whether SepF and FtsZ are sufficient for cell division in a synthetic cell remains to be seen. Due to the absence of MinC, the BMD27 strain showed aberrant polar divisions resulting in multiple minicells, which is an undesirable trait for a synthetic cell. Actinobacteria do not contain a Min system, but the reason why they have no need for such control is unclear. These bacteria grow by cell wall synthesis at their poles, whereas Bacilli synthesizes new cell wall material along their entire cell length. Maybe this growth mode is more prone to the formation of extra cell division sites next to nascent septa, necessitating the need for a regulatory Min system. E. coli and related Gram-negatives follow a comparable cell wall growth modus as Bacillus, and also use the MinCD system to prevent formation of minicells. Of course, it cannot be ruled out that Actinobacteria use other, yet unknown, systems to prevent aberrant cell division. E.g. Streptomyces use the protein couple SsgAB to assure proper localization of Z-ring formation [97]. Thus, the choice of cell wall growth might determine whether the Min regulatory system should be included in the blueprint of a functional minimal synthetic bacterial cell.

General methods

Strains and primers used in this study are listed in Tables A and B in S1 Appendix, respectively. B. subtilis strains were grown in LB medium at 37 °C and 200 RPM shaking. LB medium was supplemented with 10 mM MgSO₄ and 1% glucose in the case of the BMD strains. When required, erythromycin (1 µg/ml), spectinomycin (100 µg/ml), phleomycin (2 µg/ml) or tetracycline (10 µg/ml) were added. Chromosomal DNA for transformation to B. subtilis was prepared as described before [98]. Competent B. subtilis strains were transformed with either genomic DNA or PCR products using an optimized method described in [99].

Construction of mutants

The minimal divisome mutants were constructed using a marker-free deletion method described by Morimoto et al. [43]. Detailed construction of these and the other mutants is outlined in S1 Appendix.

Growth curves and cell length measurements

Growth curves were measured in microtiter plates, starting with overnight cultures from fresh single colonies, which were diluted to an optical density OD₆₀₀ of 0.1, grown to mid-exponential phase at 37 °C, and diluted to an optical density OD₆₀₀ of 0.05 in 150 µl.

For cell length measurements, overnight cultures were diluted to an OD₆₀₀ of 0.1 and at mid-exponential phase samples were taken, and cell membranes were fluorescently stained with 1 µg/ml FM5-95 after which cells were prepared for fluorescence microscopy analysis.

Fluorescence microscopy

Cell membranes were stained with FM5-95 (1 µg/ml final concentration, Invitrogen) and DNA was stained with DAPI (5 µg/ml final concentration, SIGMA). Cells were immobilized on 1.3% agarose-covered microscope slides. Epifluorescence images were acquired using Nikon Ti-E and the imaging program Metamorph 6 (Molecular Devices). For SIM microscopy, cell membranes were stained with Nile red (200 ng/ml, Invitrogen). To reduce background caused by excess membrane dye, the cover glass was coated with dopamine, by incubating the glass in 2 mg/ml L-dopamine in Tris-HCl pH 7.5 buffer for at least 30 min [100]. Excess dopamine was removed by washing with demi water. 1% glucose was replaced by 1% fructose when FtsZ-GFP expression was induced with xylose (1%), because of catabolite repression of the xylose-inducible promoter [101]. SIM images were taken with a Nikon Ti-E microscope and the imaging program Nis Elements AR version 4.50.

Whole genome sequencing

Chromosomal DNA was isolated as described in [102], and genomic DNA libraries were generated according to the manufacturers’ protocols using the Ion Xpress Plus gDNA Fragment Library Preparations (Life Technologies). Bar-coded libraries were prepared using the Ion Plus fragment library kit (Life Technologies) and the Ion Xpress DNA bar coding kit (Life Technologies) according to the 200-base-read Ion Proton libraries instructions of the manufacturer. Sequencing was performed on the Ion Proton system using the Ion PI Chips (Life Technologies) according to the manufacturers’ protocols. The FASTQ files were subjected to a quality control procedure, using in-house software and fastqc (www.bioinformatics.babraham.ac.uk/projects/fastqc/). Sequencing reads were mapped onto the B. subtilis reference genome (gi|255767013|ref|NC_000964.3|) using Tmap [103]. Finally, single nucleotide variants, insertions and deletions were identified using the flow-aware Torrent Variant Caller (TVC). From these variants, insertions and deletions mutations with 75% frequency and coverage of 80 were scored as relevant.

Spot-dilution assays

To test the sensitivity for 3-MBA, single colonies were cultured in LB to an OD₆₀₀ of 0.5 and serial dilutions were prepared in prewarmed LB at 37 °C. The dilutions were spotted onto LB agar plates with appropriate concentration of 3-MBA. When an inducer was required, the appropriate concentration was included in the LB agar plates. The plates were incubated overnight at 37 °C.

FtsZ-depletion sensitivity assays made use of an IPTG-inducible ftsZ gene (Pspac-ftsZ) [31]. Strains were streaked on LB plates containing 200 µM IPTG and 2 µg/ml phleomycin and grown at 37 °C. Single colonies were grown to exponential phase in LB supplemented with 200 µM of IPTG and 2 µg/ml phleomycin, then 10-fold serial dilutions were prepared in fresh pre-warmed LB. 10 µl of appropriate serial dilutions were spotted on LB plates with 2 µg/ml phleomycin and different concentration of IPTG. The presence of phleomycin was necessary to prevent excision of the Pspac-ftsZ construct, which was integrated by single crossover.

Fatty acid analyses

For fatty acid composition analysis, 500 ml cultures of B. subtilis 168 and TNVS292 (∆braB) were grown at 37 °C in LB until an OD₆₀₀ of 0.5. Cultures were quickly cooled on slush ice followed by centrifugation at 12,000 x g. Cells were washed with 2 ml ice-cold 100 mM NaCl and dry cell pellets were flash-frozen in liquid nitrogen. Samples were lyophilized, covered with argon, and stored at −80 °C. Lyophilized cell pellets were dissolved in 1.5 ml PBS and disrupted by ultra-sonication. 50 µl of the cell lysate were prepared for gas chromatography as described [104]. In short, samples were mixed with 10 nmol internal fatty acid standard and transmethylated in 1 ml 3 M HCl at 90 °C for 4 h, followed by hexane extraction. Extracts were dried under argon stream, resuspended in 100 µl hexane, and subsequently injected into a Hewlett Packard GC 5890 gas chromatograph equipped with an Agilent J&W HP-FFAP 25 m, 0.20 mm, 0.33 µm GC column. Fatty acid methyl esters were detected by flame ionization.

Membrane fluidity measurements

To assess the membrane fluidity of B. subtilis, the fluorescent membrane probe Laurdan was used [75]. Briefly, B. subtilis 168 and TNVS292 (∆braB) were grown at 37 °C in LB until an OD₆₀₀ of 0.2 and 0.8 and incubated with 10 µM Laurdan (from a 1 mM stock in DMF) for 5 min. Cells were then quickly washed four times with pre-warmed PBS containing 0.2% glucose and 1% DMF. Cell pellets were resuspended in the same buffer and OD was adjusted to 0.4 prior to transferring 150 µl aliquots to a pre-warmed black microtiter plate. Laurdan fluorescence was excited at 350 nm, and emission at 460 and 500 nm was recorded. Laurdan GP was calculated according to the formula (I460-I500)/(I460+I500).